1# Design proposal for issuing NMI on servers that use OpenBMC
2
3Author: Lakshminarayana Kammath
4
5Other contributors: Jayanth Othayoth
6
7Created: 2019-05-21
8
9
10## Problem Description
11Currently, servers that use OpenBMC cannot have the ability to capture relevant
12debug data when the host is unresponsive or hung. These systems need the ability
13to diagnose the root cause of hang and perform recovery along with debugging
14data collected.
15
16
17## Background and References
18There is a situation at customer places/lab where the host goes unresponsive
19causing system hang(https://github.com/ibm-openbmc/dev/issues/457).
20This means there is no way to figure out what went wrong with the host in a hung
21state. One has to recover the system with no relevant debug data captured.
22
23Whenever the host is unresponsive/running, Admin needs to trigger an NMI event
24which, in turn, triggers an architecture-dependent procedure that fires an
25interrupt on all the available processors on the system.
26
27## Proposed Design for servers that use OpenBMC
28This proposal aims to trigger NMI, which in turn will invoke an
29architecture-specific procedure that enables data collection followed by recovery
30of the Host. This will enable Host/OS development teams to analyze and fix any
31issues where they see host hang and unresponsive system.
32
33### D-Bus
34Introducing new D-Bus interface in the control.host namespace
35(/openbmc/phosphor-dbus-interfaces/xyz/openbmc_project/Control/Host/
36NMI.interface.yaml)
37and implement the new D-Bus back-end for respective processor specific targets.
38
39### BMC Support
40Enable NMI D-Bus phosphor interface and support this via Redfish
41
42### Redfish Schema used
43* Reference: DSP2046 2018.3,
44* ComputerSystem 1.6.0 schema provides an action called #ComputerSystem.Reset,
45  This action is used to reset the system.
46  The ResetType parameter is used for indicating the type of reset needs to be
47  performed. In this case, we can use
48  An NMI type
49    * Nmi: Generate a Diagnostic Interrupt (usually an NMI on x86 systems)
50     to cease normal operations, perform diagnostic actions and typically
51     halt the system.
52
53## High-Level Flow
541. Host/OS is hung or unresponsive or one need to take kernel dump
55   to debug some error conditions.
562. Admin/User can use the Redfish URI ComputerSystem.Reset that allows
57   POST operations and change the Action and ResetType properties to
58   {"Action":"ComputerSystem.Reset","ResetType":"Nmi"} to trigger NMI.
593. Redfish URI will invoke a D-Bus NMI back-end call which will use an arch
60   specific back-end implementation of xyz.openbmc_project.Control.Host.NMI
61   to trigger an NMI on all the processors on the system.
624. On receiving the NMI, the host will automatically invoke Architecture specific
63   actions. One such action could be; invoking the kdump followed by the reboot.
64
65* Note: NMI can be sent to the host in any state, not just at an unresponsive
66  state.
67
68## Alternatives Considered
69Extending  the existing  D-Bus interface state.Host namespace
70(/openbmc/phosphor-dbus-interfaces/xyz/openbmc_project/State/Host.interface.yaml)
71to support new RequestedHostTransition property called Nmi.
72D-Bus back-end can internally invoke processor-specific target to invoke NMI
73and do associated actions.
74
75There were strong reasons to move away from the above approach.
76phosphor-state-manager has always been focused on the states of the BMC,
77Chassis, and Host. NMI will be more of action against the host than
78a state.
79
80## Impacts
81This implementation only needs to make some changes to the system state when
82NMI is initiated irrespective of what host OS state is in, so it has minimal
83impact on the rest of the system.
84
85## Testing
86Depending on the platform hardware design, this test requires a host OS kernel
87module driver to create hard lockup/hang and then check the scenario is good.
88Also, one can invoke NMI to get the crash dump and confirm HOST received NMI
89via console logs.
90
91